The main interest of my PhD research is centered around the nested sampling algorithm. Although it was initially proposed in the field of Bayesian statistics, it has been adapted in materials science for the calculation of thermodynamic properties. Due to its computational expense, its application is to date limited to systems which can be described by simple empirical potentials. In this context, my goal is to exploit the accuracy and computational efficiency of recent neural network force fields by using them as a backend for nested sampling. This opens up the route towards thermodynamics simulations of more complex materials like metal oxides close to ab-initio precision.